Authentication of Wine by 1H-NMR Spectroscopy: Opportunities and

Chapter 6

Authentication of Wine by 1H-NMR Spectroscopy: Opportunities and Challenges Downloaded by MONASH UNIV on November 25, 2015 | http://pubs.acs.org Publication Date (Web): November 24, 2015 | doi: 10.1021/bk-2015-1203.ch006

Susanne Esslinger, Carsten Fauhl-Hassek, and Reiner Wittkowski* BfR Federal Institute for Risk Assessment, Department of Safety in the Food Chain - Max-Dohrn-Str. 8-10, D-10589 Berlin, Germany *E-mail: [email protected]. Phone: +493018412-3376.

The verification of wine identity and authenticity is of urgent importance in the current context of a growing market globalization. As a result, wine authentication is an indispensable as well as essential aspect in today’s consumer protection. Regarding its chemical analysis the matrix wine is challenging, whereas its valuable characteristics are based on different factors, such as different types of tastes, the geographical origin associated with the growing conditions, vintage and grape variety. Accordingly, the range of analytical methods to enable a comprehensive characterization of these products is highly diversified. In this context, 1H-NMR spectroscopy is currently employed to characterize wine in terms of targeted as well as nontargeted analysis in only a few minutes and therefore allows the simultaneous investigation of diverse parameters. As the targeted approach enables an identification and quantification of different key ingredients in wine, the nontargeted, also called fingerprinting analysis, with subsequent statistical data evaluation investigates the whole spectrum of the matrix. Therefore, the capability to detect known adulterants, but also the ability to detect further abnormalities, and the assessment of challenging authentication parameters (grape variety, origin, vintage) appoint this technique of utmost interest for quality control, research and control institutions. Particularly concerning the nontargeted approaches, current relevant scientific literature is typically based on feasibility, demonstration or research studies within one laboratory on one instrument exclusively, which © 2015 American Chemical Society In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

restricts validation possibilities and degrees. Validation of the whole analytical procedure including statistical data evaluation and consistency of the measurement over time, instruments and laboratories is, however, essential for routine application and in official control. Therefore, the use of nontargeted fingerprinting approaches is due to actual missing validation strategies still restricted for official control purposes, thereby offering complex challenges for the scientific community.

Downloaded by MONASH UNIV on November 25, 2015 | http://pubs.acs.org Publication Date (Web): November 24, 2015 | doi: 10.1021/bk-2015-1203.ch006

Introduction Usually, appearance, aroma, taste and mouth feeling are attributes, defining the value of a wine or respective products and effecting the consumer’s acceptance, even if only unconsciously. In addition to these sensory properties, there are other product criteria, including brand, labeling of ingredients and their quantitative data, as well as the geographical origin, which are crucial in terms of the consumer’s decision to buy. Not at least because of the increasing globalization of supply chains, the consumers are aware on the authenticity of a product in an increasing degree. Therefore, the requirements on food, the appropriate ingredients and their labeling are strictly regulated in the European Union (EU). In this context, the term authentication describes the confirmation of all requirements regarding the legal product description or the detection of the fraudulent statements (1, 2) particularly in view of: (i) the substitution by cheaper but similar ingredients, (ii) extension of food using adulterants (e.g. water, starch including exogenous material) or blending and/or undeclared processes (e.g. irradiation, extraction), (iii) the origin, e.g. geographic, species or method of production. The increasing requirements on wine authentication (not at least because of the complex composition of such products), result in the need for reliable strategies in the wine control. Actually official wine control in Europe consists of the: • • •

control of wine quality/fair merchantable quality: sensory and “off-flavors”, control of statements on the label, e.g. alcohol content, quality, grape variety (blending), geographical origin, vintage, chemical adulterations, e.g. illegal acidification, addition of water, glycerol, alcohol, sugar, dyes, sweeteners, preservatives, flavors.

In view of these aims, the classical authenticity assessment of wine is usually based on the analysis of specific marker compounds, which are indicative for a certain property of the product. Wine authentication is therefore performed routinely (in official food control) by targeted analysis using the classic 86 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


wet-chemical approaches, e.g. stable isotope analysis by isotope ratio mass spectrometry (IRMS), site-specific natural isotopic fractionation (SNIF)- NMR® spectroscopy, Fourier-transform infrared spectroscopy (FT-IR, e.g. (3)), high performance liquid chromatography as well as gas chromatography (4). In the last few years the nontargeted analysis, also called food fingerprinting, obtained increasingly importance (5), but is not yet established in official wine control. These applications are usually based on spectroscopic and spectrometric data providing the capability for a comprehensive characterization of the investigated matrices, the differentiation of the samples due to their botanical or geographical origin, the production process (e.g. organic versus conventional), the identification of adulterations etc. This strategy is based on the principle of metabonomics, describing the scientific study of small molecules, the metabolites, of a biological system based on comprehensive chemical analysis (omics technologies) with the aim to detect as many substances as possible (6). Because of its up-coming importance in wine authentication the general steps of nontargeted analysis are highlighted in Figure 1.

Figure 1. Schematic overview on the general steps and their applicability of nontargeted analysis. In principle, this approach in food and wine analysis is characterized by a nontargeted, fast and easy spectroscopic or spectrometric analysis, acquiring sample specific fingerprints. These profiles are compared to a large database of authentic or typical samples using a univariate and/or multivariate statistical approach. Essential part of the statistical evaluation procedure is the data pre-processing, whereby data are often prepared, in order to transform them into a suitable form for statistical analysis. Generally, the choice for the different possibilities in data pre-processing depends strongly on the used analytical technique as well as the objective (1). The objectives which are pursued in the subsequent multivariate data analysis divide the respective applications in three main categories (7): 87 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


• • •

data description (explorative data structure modeling), discrimination and classification, regression and prediction.

Afterwards, in dependence of the used analytical method and the statistical evaluation, it might be possible to identify compounds responsible e.g. for the differentiation or classification. Therefore, the nontargeted approach is also called a bottom-up strategy. Within a comprehensive literature review by Esslinger et al., where food fingerprinting studies of the last five years were analyzed, it was stated that the spectroscopic methods Near Infrared (NIR), Fourier-transform (Mid) Infrared (FT-(M)IR) and Nuclear Magnetic Resonance (NMR) spectroscopy were the most common techniques used in food fingerprinting (1). Compared to this, mass spectrometry (MS)-based methods were less often used, thereby applying various kinds of techniques such as direct MS (Fourier-transform ion cyclotron resonance, time-of-flight, secondary electrospray ionization and proton-transfer-reaction mass spectrometry) or MS coupled to different chromatographic separation techniques (Gas Chromatography (GC), Liquid Chromatography (LC)) as well as the direct injection (1). As the following article focuses on the authentication of wine using 1H-NMR spectroscopy, the topic will be discussed accordingly and from the perspective of the official wine control, illustrating the possibilities and challenges of the technique.

NMR Spectroscopy Typically, NMR spectroscopy is used as tool for structure elucidation in organic chemistry. Super conducting magnets with very stable high magnetic fields are indispensable tools in structure elucidation since decades. The strength of the magnetic field is either expressed in Tesla (T) or by the corresponding resonance frequency of the protons in Megahertz (MHz). Apart from its function as structural elucidator NMR spectroscopy became and becomes a more and more interesting tool in metabolomics (e.g. (8, 9)), nutritional (e.g. (10, 11)) and food science (e.g. (12)) also due to its unique quantitative properties. In relation to wine analysis the so-called SNIF-NMR® spectroscopy is well established for the detection of illegal sugar addition to must and wine (chaptalization) for more than 25 years. In fact this measurement consists in the site-specific quantification of deuterium (D) in ethanol, which is indicative for the type of the initially fermented sugar (13, 14). 13C-NMR spectroscopy was used by the working group of Professor A. Rapp in Germany for the quantification of wine ingredients by NMR in the 1980’s (15–17). However, this methodology never found application in the routine analysis of wine. Recently - during the last 10 years – quantitative 1H-NMR spectroscopy for fruit juice and even more recently for wine became of interest in research but also in routine applications. Instrumental developments enable the fast and reliable quantification in combination with easy to handle NMR spectrometers 88 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


(“push button”) of wine ingredients as well as authentication of wine by the application of chemometrics/statistics to 1H-NMR data. 1H-NMR spectroscopy has a number of advantages in comparison to 13C-NMR spectroscopy and other techniques but also some limitations. 1H-NMR is more sensitive – based on its higher gyromagnetic ratio, higher natural abundance, compared to 13C-NMR (18). However, due to its low spectral resolution, signal overlap occurs in 1H-NMR spectroscopy and needs to be considered carefully in the spectra evaluation. In addition, it must be clearly noted as general remark that 1H-NMR is not and will not be a tool for trace analysis, but the limits of detection for quantification go down to the low mg/L range for routinely used NMR instruments of the newest generation. The sensitivity depends of course on the field strength and the type of probe used (19). Typically in case of food and wine analysis instruments of 400 MHz up to 600 MHz are established (20–22). NMR spectroscopy is characterized by an excellent linearity; the generated signals are proportional to the underlying concentrations over orders of magnitudes (19). NMR data contain also structural information as it is used for structure elucidation, meaning that unknown signals for example might be assigned to certain compounds. Data acquisition is done in a few minutes with a reasonable signal to noise ratio including the detection of minor components. As general drawback the costs of instrumentation needs to be mentioned.

Wine Authentication by 1H-NMR – Status Quo Sample Preparation As result of the mentioned advantages and technical progress a lot of scientific literature was published investigating the applicability of 1H-NMR spectroscopy for wine authentication (23–26). Starting with the sample preparation of the wine matrix, several approaches have been described. Most studies investigating this topic, take typically a certain amount of the wine sample as it is (e.g. (24)), eventually filtration was performed. Apart from the simple addition of deuterium oxide (D2O), which is necessary as “lock substance”, buffer systems containing D2O and often phosphate have been used for the 1H-NMR analysis of wine, resulting in more stable and reproducible spectra. This application of phosphate buffer systems is already known and common practice in the metabolite profiling analysis of biofluids, e.g. urine (19) aiming the reduction of the pH-dependent chemical shift variation in 1H-NMR spectroscopy. Son et al. report lyophilization as sample preparation with the benefit of getting rid of the water and ethanol signals but on the expense of a more laborious sample preparation (26). In order to adjust the chemical shift scale the addition of trimethylsilyl propionic acid (TSP), the water soluble correspondent to tetramethylsilane, is very often used. For quantification purposes – this is a different aim than the chemical shift adjustment – also other internal standards, e.g. anthracene (27, 28) or nicotinamide (29) have been described. Furthermore, a practical application using 1H-NMR spectroscopy (400 MHz) in the field of wine analysis and authentication was developed and is commercially available. Joint efforts of Bruker BioSpin GmbH and SGF International e.V. 89 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


resulted in a NMR-based screening method, so-called FoodScreener™ for wine (WineScreener™) as well as for fruit juices (JuiceScreener™), combined with the Profiling™ technique (30–33). Within this concept, it is possible to compare actual nontargeted spectral 1H-NMR data with the corresponding group of reference spectra (e.g. database of several thousands of reference juices, obtained from production sites all over the world) using verification models. The aim of the classification or verification analysis in the case of the WineScreener™ is the determination of the grape variety, geographical origin and vintage. Furthermore, the models enable the detection of any deviation from authentic reference data in the nontargeted approach. Besides a nontargeted analysis of matrix, this screening method also provides targeted results in a single measurement. Especially in wine analysis, the targeted evaluation for varieties and origins provides automatic quantification of about 30 parameters per sample. By the way for fruit juice (JuiceScreener™), currently the quantification of about 60 parameters is offered in the commercial tool. In case of the analysis of wine, the sample preparation of the Bruker WineScreener™ protocol consists of the addition of special buffer solution containing D2O and afterwards the pH of the wine buffer solution is carefully and very precisely adjusted by the addition of hydrochloric acid (HCl) or sodium hydroxide solution (NaOH) against a reference solution (Bruker). After the adjustment an aliquot of the solution is transferred into a NMR tube and submitted to the 1H-NMR data acquisition. A general aspect of the sophisticated NMR application (WineScreener™) is that this system shows a very high sensitivity, so that e.g. the filling height of the solution in the tube as well as the quality of the tubes play an important role. Here, e.g. the wall thickness is affecting the amount of sample in the NMR measurement coil. On the deuterium channel of the NMR instrument the magnetic field is locked and homogenized (“shimmed”). The proton channel is used for data acquisition. 1H-NMR

Measurement

The next step in the analysis of wine by 1H-NMR spectroscopy is the actual acquisition of the NMR spectrum. Due to the composition of wine its 1H-NMR spectrum is predominated by the water signal (~4.8 ppm) followed by the ethanol signals, the quartet of the methylene group and the triplet of the methyl group at 3.6 ppm and 1.2 ppm, whereas further signals are not visible at first glance in a simple 1H-NMR spectrum. In NMR spectroscopy water suppression is applied in routine and in an excellent reproducible matter but in wine apart from water the ethanol is present. Thus multiple signal suppression was introduced in which the water signal plus the ethanol methylene and methyl group signals are suppressed by a so-called eightfold suppression ((34, 35), Figure 2). This routine has been perfectly integrated into the automation of spectra acquisition and results in the significant signal enhancement of the minor components. This development has been also integrated in the WineScreener™ measurement procedure. The distortion of the spectrum sections, close to the suppressed areas, remains minimal, thus signals very close to the suppressed range can be evaluated and quantified as well. 90 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


Figure 2. Presentation of the effect of eightfold suppression of water and ethanol signals in a wine sample (according to (34)), showing that the signal intensities of remaining minor wine ingredients are increased. Particularly, in case of 1H-NMR analysis of wine, pH effects for some signals, mainly for organic acids but also for other signals, were observed. As already mentioned in the previous chapter “Sample preparation”, the chemical shift of certain signals varies with the pH value of the wine sample, but this phenomenon is unpredictable so far. The pH value of wine typically varies between 2.8 and 3.4 and in fact wine itself must be considered as buffer system, due to its high content of minerals and acids. These described signal shifts cause difficulties in the subsequently performed statistical spectrum analysis, either in quantification or in multivariate statistics. For quantification particularly in automation the integration routine is often fixed to particular limits of the signal and any shift could cause miss-integration. Multivariate statistics also depend on stable chemical shifts in the NMR data, otherwise these evaluations might end up in erroneous classifications led by the wine’s pH value instead of the actual considered attribute for example. The physical pH adjustment by adding acid or base is not the only way to accommodate these shifts; it can also be done mathematically by application of certain algorithms, e.g. Interval-Correlation-Shifting (ICOSHIFT (36);). A further major breakthrough in NMR analysis of food and here wine was the improvement of the repeatability/reproducibility of the acquisition. In Figure 3 the overlay of 21 spectra of different preparations of one “quality control sample” (white wine) is shown. Developments in the hardware, such as improved stability control of temperature, improvements in the phase and baseline stability and their adjustments in automation contributed to such a prominent repeatability achievable on modern NMR instruments. This data consistency and stability – in longer term view – is “the prerequisite” for further developments in the field of authentication e.g. for the creation of one common spectral database which goes behind the requirements of typical “metabonomics” application. In metabonomics 91 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


studies usually differences in the metabolites composition between samples are considered at a certain stage of time – when the measurement is performed – whereas authentication requires essentially data consistency and measurements over longer periods of time. As recently shown by Minoja and Napoli excellent reproducibility is achieved with the Bruker WineScreener™ also at different sites employing different instruments of the same type and vendor (37).

Figure 3. Overlaid 1H-NMR spectra of 21 replicates of a “quality control sample” (white wine), showing the good reproducibility of the technique.

Targeted Data Evaluation Subsequent to the sample preparation and analysis by 1H-NMR, the evaluation of the acquired data takes place, whereas the quantification of ingredients is the next step, in case of targeted analysis. Very different procedures are applied for the quantification of wine ingredients by 1H-NMR spectroscopy: use of internal standards, external/matrix calibration and standard addition are known and established quantification methodologies in classical analytical chemistry. Further, multivariate regression models, e.g. partial least square analysis (PLS) with reference data can be used. Herein, an external matrix or parameter – typically the concentration of the wine ingredient in question – which has been determined by an external procedure, e.g. the reference method, is correlated with the NMR spectra. Signals or areas of signals which correlate will be identified and used to calibrate the multivariate model. There are also further and more NMR specific quantification procedures described, for example the so-called PULCON (pulse length based concentration determination) method (38). Due to the excellent measurement stability of 92 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


modern NMR instrumentation, it is possible to determine the area per µmol proton using externally measured chemical standards (in an extra tube). A quantification of the analytes in the actual sample is then conducted by using this constant response factor and enables even the quantification of wine ingredients without any respective standard available. However, quantification of wine ingredients by 1H-NMR spectroscopy requires deep understanding of the underlying molecular structure and great care during the integration. Therefore, 1) relevant signals need to be assigned properly, 2) integration of the signal areas needs to be done with great care including the reflection of possibly overlapping signals, which might require a curve fitting procedure (e.g. deconvolution) for integration, and finally, 3) over- or underestimation effects by e.g. pulsation transfer or incomplete relaxation of certain signals of particular analytes need to be proven by extensive spiking experiments. Nontargeted Data Evaluation The nontargeted evaluation of the acquired spectral data needs, in most cases, dedicated statistical data pre-processing to compensate differences, e.g. slight signal shifts, unequal baselines etc.. The most widely used mathematical approach to reduce the acquired data size and to minimize peak shift effects in NMR spectroscopy for multivariate statistics is the so-called bucketing (or binning) (1). Bucketing is based on segmenting a spectrum into small areas (buckets/bins) and taking the area of the spectrum for each segment for further evaluation. This procedure results in a new spectrum, containing a significantly reduced number of data points. Experiments investigating the authenticity of international wines (Esslinger, Strassberger, Blaas, Fauhl-Hassek, 2013, unpublished, data not shown) showed that a major drawback of bucketing might be the loss of a considerable amount of information compared to the original spectra. Additionally, in some cases, the borders of the buckets are fixed and applied rigorously in automation, irrespective of potential deviations. By that, a non-correct alignment can lead to erroneous bucket loads. Larger signal shifts between different spectra, e.g. due to instrumental variations in the analysis, pH value dependent variations or changes of the salt concentration in the matrix, may lead to a larger variation in the resulting data set/matrix and obscure the identification of patterns in the data set. Subsequent to the sample pre-processing, e.g. bucketing, statistical data evaluation is performed. For sample verification within the commercial application WineScreener™ the whole NMR profile of a specific sample is compared with the corresponding group of reference spectra (database (39),) at first hand explorative and univariate. For this, quantile plots are generated to visualize the respective results. By visual and/or mathematical inspection of the “fitting”, this approach enables to determine deviations from the “reference data”. The inspection of the quantile plot, an example is given in Figure 4, is similar to the classical authentication process considering single parameters transposed to the whole spectral information. 93 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


Figure 4. Quantile plot of 353 white wine samples (presented in Table 1), analyzed by 1H-NMR (1.2 - 3.7 ppm); red line represents the median value; further colors are the distribution of the sample’s variation. Quantile plot was generated using a MATLAB (MathWorks®) code.

In case of classical single parameter assessment, the approach consists in determining the concentration of selected natural components which are characteristic for the specific type of wine. The comparison of the data obtained with the previously established normal concentration ranges of the substance is the particular decisive factor of this assessment. The point to start with is the collection of data of authentic or unsuspicious samples. Here, authentication data of “unsuspicious samples” are collected. From these data a calculation of the authenticity range is performed usually by employing the student factor, which is a tabulated value for a certain probability and number of observations (40). This factor is multiplied by the standard deviation observed giving a so-called confidence interval which represents the “authenticity range”. This authenticity range for single parameters is in principle similar to the quantile limits in the spectral evaluation. The next step is the comparison of the actual measurement value with the authenticity ranges. If the value fits into this range, it is unsuspicious and the sample or at least the parameter investigated is assessed to be compliant. On the other hand if the value lies outside, it is a clear indication for fraud and further investigations or actions should be carried out (41). After the NMR measurement, followed by the statistical data pre-processing, the resulting data matrices comprise partially (in dependence of the chosen data pre-processing procedure) of several thousands of variables including analytical results as well as meta data (e.g. variety, geographical origin, vintage). Figure 5 shows a respective example of a data matrix after the analysis of wine by 1H-NMR spectroscopy. 94 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


Figure 5. The exemplary presentation of a data matrix after NMR analysis, including metadata, where the samples are listed line-by-line and the variables (analytical results and meta data) are mentioned column-wise.

In this case, the observations (samples) are listed line-by-line. The first columns on the left side of the table describe the respective metadata, followed by the columns including the analytical results. This matrix is fundamental for the subsequently performed multivariate data evaluation. Multivariate statistics can generally be divided in unsupervised and supervised classification methods as well as in regression methods (for quantification). Unsupervised methods aim to identify patterns in the data that could not be derived from a priori available knowledge of the data. Several tools for exploring the data are available. Principal component analysis (PCA), factor analysis (FA) and hierarchical cluster analysis (HCA) are variable reduction techniques defining a number of latent variables by making linear combinations of the original variables following a given criterion. For all methods, the projections of the n objects from the original data space on a latent variable are called the scores on this latent variable (42). In contrast, supervised methods use calibration or training sets with a priori known information (e.g. about variety or vintage) to build a classification model. Then, the model is tested using an independent sample set also with a priori known information to validate the predictive properties of the model before using it on unknown samples (43). The most popular supervised techniques for the classification include linear discriminant analysis (LDA), soft independent modeling of class analogy (SIMCA), partial least squares–discriminant analysis (PLS-DA) and orthogonal projections to latent structures–discriminant analysis (OPLS-DA). As already mentioned, regression methods are used for multivariate calibration to quantify/predict properties of interest such as total flavonoid content, anti-oxidant activity, etc. 95 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

Excellent and very promising results have been reported by applying 1H-NMR spectroscopy for some of the classical challenges in wine analysis: the proof of the grape variety and vintage. Apart from the assignment of the geographical origin these attributes are still of utmost interest in the official control of wines. Excellent differentiation between different white wine varieties was achieved by 1H-NMR analysis (31). Referring to the variety authentication, two examples of targeted approaches to demonstrate what currently is done and possible in official wine control should be mentioned.


a)

Anthocyanin pattern Although similar types of anthocyanins are found in different grape varieties, the relative amounts of the individual compounds differ. For example, it has been noted that Pinot Noir grapes contain no acylated anthocyanins. These compounds have proven to be particularly characteristic for certain grape varieties, with considerable practical significance (44, 45). This feature of Pinot noir wines is successfully applied for their variety control in Germany, and the differentiation of Merlot wine would be straight forward. b) Content of shikimic acid In addition, for some questions the shikimic acid gives interesting information on the authenticity of the wine variety, here again the Burgundy wines including the white wine varieties show specific concentrations. For example Riesling wines are characterized by a high content in contrast to the Burgundy wines, which show a low concentration of shikimic acid. Therefore, shikimic acid is an indicator for certain varieties and can also be indicative for some others. According to the information of the German wine control the number of objections dropped down drastically after the consideration of the confidence limits for the Burgundy wines. However, these are the only forensic applications in the targeted variety authentication of wine. Certainly there are many further open questions – concerning this topic – including red wine varieties, differentiation of Merlot/Cabernet Sauvignon, Syrah etc. and white wine varieties, e.g. Sauvignon Blanc/Chardonnay/Riesling/Silvaner/Müller-Thurgau. These questions regarding the variety proof actually can not be answered by targeted analysis. In addition, the indication of the vintage, of very specific geographic origins and a single variety was demonstrated by using 1H-NMR spectroscopy by Godelmann et al. (31). It should be noted that there is no forensic methodology for the proof of the vintage, and therefore these results are very interesting for official wine control. It may be mentioned in this context that the measurement of the 14C in ethanol has a resolution of about five years only and the activity after the extended nuclear tests in the mid sixties dropped down to the original natural level, which means that no effect can be measured nowadays. Particular the “reserva” attributes with different obligations of storage - are of interest. Sometimes doubts about the 96 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

history of certain wines occur, by suspicious sensory analysis and/or anthocyanins pattern analysis (decreasing with ageing, polymerization), but no forensic method of analysis for vintage verification could be established in routine/control so far (46).

Further Research Using 1H-NMR Spectroscopy for the Variety Authentication of Wine


Experimental Design In order to tackle the variety authentication, further research was conducted (Esslinger, Strassberger, Blaas, Fauhl-Hassek, 2013, unpublished). The aim of this study was the development of a 1H-NMR based method for a nontargeted analysis of wine to investigate the matrices with regard to authenticity. For this a set of 495 commercial wine samples was investigated. Four different red wine varieties and four different white wine varieties were considered: Merlot, Tempranillo, Pinot Noir and Syrah for the red wines and Riesling, Chardonnay, Sauvignon Blanc and Silvaner for the white wines (Table 1). The study was conducted intentionally with commercial wines.

Table 1. Overview on the Investigated Wine Samples and Respective Number of Samples (n) within the Presented Study Investigating the Variety Authentication of Wine by 1H-NMR Spectroscopy and Multivariate Statistics Red wine variety

n

White wine variety

n

Merlot

46

Riesling

206

Tempranillo

45

Chardonnay

49

Pinot Noir

22

Sauvignon Blanc

48

Syrah

29

Silvaner

50

In this context, it should be mentioned, that one of the main criticism of authenticity testing in general is that the data used as reference are not covering the natural diversity and that possibly some effects – biological or oenological – on the parameter in question have not been adequately investigated. Authentic samples or experimental samples (e.g. from the stable isotope databank according to Regulation (EC) 555/2008; European Union, 2008 (13)) of which the authenticity is “guaranteed” provide the risk that their production might differ – even only slightly – from commercial samples, and therefore may not reflect the reality sufficiently. A strong argument for using commercial samples is that all possible effects and variations are covered already and the data are therefore 97 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


fairly robust. One the other hand natural differences may be caused by authorized oenological practices, which may result in variances in the acquired spectral data, that make the statistical evaluation more difficult. However, within the presented study, commercial wine samples were used to establish a respective sample preparation including a pre-processing of the acquired data. The samples were analyzed using a NMR spectrometer (400 MHz). The sample preparation was performed according to the WineScreener™ protocol but with slight deviations (47, 48): The set of samples was prepared after the addition of a laboratory internally developed phosphate buffer including the additional pH adjustment to a value of 3.00 ± 0.04. The achieved data set was used to investigate different pre-processing methods: → bucketing (including different techniques as well as different bucket widths), → identification and elimination of outliers using PCA, → classification according to grape variety using PLS-DA.

Outlier Identification Since the multivariate data analysis is based on the variance in the data space, a check with respect to the identification of outliers must be carried out in a first step. The visual inspection of all acquired spectra was performed, but only hard outliers can be detected easily as shown in Figure 6A. All spectra are overlaid (black lines), whereas the red line spectrum deviates. In this case the water suppression during the acquisition did not perform perfectly, thus this sample was re-measured. Other reasons for outlying spectra have been identified, e.g. erroneous pH adjustment or the spoilage of a sample. These types of outliers are also easily detected by applying PCA, which is typically performed to obtain an overview on the acquired data as exploratory analysis. In Figure 6B a PCA scores plot is shown, visualizing the observed outliers (red dot).

Figure 6. Two examples of determining outliers after NMR analysis. A: Overlaid 1H-NMR spectra, indicating one outlier (red line). B: Outlier identification by the use of exploratory PCA analysis. The outlier is highlighted in red. 98 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


Quality Assurance Quality assurance measures e.g. the inclusion of quality control samples are very important also in nontargeted analysis and should become standard. The implementation of quality control samples serves as control of the whole performance of the approach (including the sample preparation and NMR measurement) with regard to its repeatability and variance. Moreover, the quality control (QC) sample should have a certain shelf life and thus be stable over the observation period. Compared to targeted analysis, where only the analyte or a group of compounds of interest has to be free of decomposition or degradation, in nontargeted analysis, the stability of the matrix itself has to be ensured. This is difficult because most foods are subject to alteration and spoilage over a certain period of time (1). In general, it must be noted that the basic prerequisite for any reliable mathematical model is that the variance of the QC sample replicates (sample preparation and measurement) must be smaller than the natural variance of the “authentic” samples, as demonstrated in this PCA scores plot in Figure 7.

Figure 7. Exemplary presentation of a PCA scores plot of investigated wine samples, including the quality control (QC) samples, showing their small variance compared to the other “authentic” samples. Furthermore, using a QC sample, which should be of a similar composition as the investigated authentic samples, and exploratory analysis (e.g. PCA) a possible time dependent drift might also be detectable. As an example, Nietner et al. investigated feed material using nontargeted FT-IR spectroscopy and used starch, which is a main component of the investigated matrix, as QC sample (49). Within the presented investigations of wine samples, a commercial white wine was used as QC sample, which was prepared and analyzed twice a day. The subsequent evaluation of the acquired spectral data focused on the performance of the technical equipment (NMR), e.g. with regard to the performed water suppression and the used shim quality as well as the sample preparation, e.g. imprecise pH value adjustment. For this, several parameters (integral of malic acid and glucose signal, chemical shift of tartaric acid and TSP, as well as the peak width at half-height of the TSP signal) were used to create Shewhart control 99 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

charts. As an example the “Upper Control Limit - Lower Control Limit”-range of the chemical shift of tartaric acid was calculated at 4.625 ppm ± 0.008 ppm. In case of exceeding these ranges, the causes have to be found and removed. In a second step, the multivariate evaluation of the analyzed control samples (as mentioned in this chapter) took place.


Data Evaluation After identification of outliers and monitoring the QC sample, a further PCA was performed to get an overview on the present data set. The respective scores plot (Figure 8) representing the explained variance of the data, shows the entire data set in multivariate space.

Figure 8. PCA scores plot (over the first two principal components), calculated with all analyzed wine samples, colored by variety, showing a grouping of red wines (right side) and white wine (left side). Within this model, the data points have been colored according to the variety of the respective wine samples showing a clear separation of red wine (right group) varieties and white wine varieties (left group). Therefore, the main variance of the data set is based on the differences between red and white wine. The differentiation between these two groups is already of high interest due to notifications about the illegal addition of white wine to red wine as well as decolorized red wine to white wine in dependence to the market/consumer preference. Particularly, the addition of small proportions of white wines to red wine is challenging for the analytical chemist. Perceptively, the development of a decision cascade/tree might be of interest, e.g. the differentiation between red and white wine, followed by the possibility to distinguish between several varieties. Thereafter, objectives like geographical origin and vintage might be of interest. 100 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


Subsequent to the exploratory analysis of the data set (by PCA), in many cases, a classification/regression model was developed. In order to avoid over-fitted models, the appropriate validation of the mathematical models is essential in the application of multivariate statistics. Only the validation enables the reliable and sustainable development of models to be used in control. One fair option for doing so is the creation of two data sets, one training set, comprising of e.g. 2/3 of the analyzed samples which is used to generate the classification model and one independent test set of e.g. 1/3 of the data to proof the prediction ability of the model. This approach is called “external validation”. Within the presented study, the total data set, consisting of about 500 wine samples, was divided into two separate models of white wine varieties and red wine varieties, respectively. Afterwards, PLS-DA was applied as classification method to the training data set. Due to the fact, that PLS-DA is a supervised method, different classes were generated in each model, each for one variety. The application of these models to the test set resulted in correct classification rates between about 70% and 98%. Although the commercial application WineScreenerTM reports better classification results on the similar varieties, our results appear encouraging for further investigations.

Future Challenges Validation Procedures The validation of an analytical method serves the “confirmation by examination and the provision of objective evidence that the particular requirements for a specific intended use are fulfilled” (European Standard EN ISO / IEC 17025: 2005; International Organization for Standardization, 2005 (50)). This allows preparing the operational readiness of the developed method in routine analysis as well as its ensuring by statistically based quality assurance measures. With regard to the preventive consumer protection, it is necessary/essential for laboratories of official food control that non-standardized or in-house developed analytical methods are validated to confirm that the methods are suitable for the intended use. The validation of these methods comprises the complete analytical procedure from e.g. the sampling or sample extraction to statistical evaluation of acquired data. Its extend should be set according to the need to fulfill the requirements of the intended use or the relevant application area. Common validation parameters are e.g. the determination of the linearity range, the limits of detection and quantification, accuracy, precision, trueness, robustness as well as the calculation of the measurement uncertainty. In few areas, e.g. pesticides or pharmacological drugs, requirements and parameters to validate respective analytical methods are precisely defined. Validation of nontargeted approaches is neither defined nor scientifically agreed, classical terms such as: limit of detection, limit of quantification, accuracy, trueness, precision, measurement uncertainty do not fit. Therefore, new validation strategies are needed for nontargeted fingerprinting, completely new 101 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


concepts required, including the definition of the relevant questions. So-called “good practice” procedures for the model validation of nontargeted methods are needed, on the conduction itself but also for the publication of such studies. With special regard to this latter point, parameters and indicators need to be determined and stated when published, e.g. the principal mention of the analyzed number of samples, the used validation procedure or the size of the final matrix, which was used for the development of the mathematical model. The scientific community has realized these needs and certain activities e.g. in the similar field of metabolomics started (51, 52).

Detection of the Unknown

A general point of discussion is the ability to detect known but also unknown adulterations using the respectively established mathematical model. Typical multivariate statistics (e.g. DA, LDA, PLS-DA, SIMCA) perform better with a previous training (by internal or external validation) of the adulteration of interest. That implies that the respective deviation from the product is already known. Using these trained models to detect currently unknown adulterations that have not been trained, one may run into danger to get false-positive or false-negative results. Further investigations are needed to identify and establish mathematical models which enable the reliable detection of known and unknown deviations from a typical food product. Some approaches try to resolve this particular problem – detecting the unknown – by univariate data evaluation, e.g. quantile plot, z-score, or by single parameters derived from the multivariate statistical process control, e.g. Mahalanobis distance, Distance to the model of X-space (DModX). Zhang and Nie exclusively used the Mahalanobis distance to classify the adulterated samples of Radix Astragali (Chinese medicine) as acceptable or unacceptable for the known data set (53). This tool is similar to the Euclidean distance, but takes into account that some variables may be correlated and therefore, measure more or less the same properties (54). A further technique to detect adulterations (moderate outliers) is by consideration of the model residuals (DModX). DModX describes the distance of the observation to the X model plane ore hyper plane and is also known as the residual error or the residual standard deviation (7, 55). The evaluation of the results is similar to the Mahalanobis distance approaches. For this, evaluation criteria are to be defined, e.g. the maximum value of DModX for the unadulterated/authentic samples is to be considered as threshold above which a test sample could be considered as suspect. The investigations of chemical contaminations in carbonated soft drinks (by NMR spectroscopy (55);) as well as of melamine adulterations of soya bean meal (by NIR (56);) showed that the identification of moderate outliers using DModX parameter from a SIMCA model is particularly suitable for the detection of contaminations without previous knowledge of their identity. 102 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

In general, multivariate data analysis of fingerprints has some advantages compared to univariate statistics of e.g. single components of the fingerprints (57, 58). Advantageous in first instance is the fact that the whole spectra information detected with the nontargeted approach is used. Nonetheless, also univariate statistical approaches were used to evaluate the data sets. Despite these alternatives, further investigations are needed to identify and establish mathematical models which enable the reliable detection of known and unknown deviations from a typical food product including wine.


Data Consistency and Exchangeability

Data consistency is a very important issue in nontargeted authenticity studies, including NMR methodologies. The typical procedure in case of nontargeted NMR studies as well as in other nontargeted studies consists of several steps: the method development (including sample preparation, measurement protocol, standardized data processing) and evaluation takes place in one laboratory (the usual situation is that only one high-tech e.g. NMR instrument is present), ending up in a “measurement procedure”. By this kind of procedure a defined set of samples is analyzed and evaluated. As mentioned earlier the validation of the procedure should be conducted according to the “good practice” which is to be defined. In this relation only in some studies quality control samples are analyzed with the set of samples and considered in the process to monitor the consistency and stability of the whole procedure over the measurement period. Also only in some occasions the model is extended by the later measurement of further samples (e.g. in a second batch at different point of time), sometimes this important fact in view of data consistency is simply not stated. Considering the “one laboratory” approach, from a more global point of view it should be mentioned that many research studies and publications end with the measurement of one series of samples, often obtained within a project. Although very interesting results are obtained, only a few studies include the measurement of more batches. However, what was introduced with Bruker WineScreener™, is a proprietary measurement procedure, what means that a second laboratory and also further laboratories, equipped with the same instrument, can use the data together. This fact must be acknowledged and according to the author’s present knowledge this is very unique in the area of nontargeted analysis. Although considerable efforts and results in this field, this theme needs to be further strengthened and further investigations should be conducted, especially in view of the application of the nontargeted analysis and its comparability after using analytical instruments e.g. from different vendors. Especially with regard to the establishment of these approaches in the official wine control, it is to be considered that standard setting organizations such as Codex Alimentarius and the International Organisation of Vine and Wine (OIV) adapt proprietary methods only, if specified prerequisites are given. 103 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

Conclusion


1H-NMR

has a lot of very interesting capabilities. It is fast, the quantification of many wine ingredients is possible within one analytical run resulting in a cost efficient analysis, substituting other methods of analysis (HPLC/GC etc.). These advantages are overshadowed by the high investment costs for the analytical instrumentation. Besides the high linearity range of this method, it should be mentioned that the NMR spectroscopy is not able to detect compounds in lower concentrations (trace analysis). But the classical wine analysis/control is focused only partially on trace compounds. In addition, the quantitative NMR analysis is a so-called primary reference measurement procedure, enabling the quantification of compounds without relation to (similar) standards. The nontargeted analysis by 1H-NMR can potentially serve the authentication of wine samples with regard to the discrimination between e.g. varieties, vintages or geographical origin, in complementation of established targeted methods. Furthermore, there are indications that the NMR analysis may allow the accomplishment of so far open challenges in wine control, e.g. the investigation of the labeled vintage. Nevertheless, the forensic application in official control requires the appropriate validation of such a nontargeted methodology. Standard procedures have to be identified and agreed scientifically as discussed in this contribution. In addition, further concepts for the use of common data bases need to be developed. Particularly the data sharing/ownership/maintenance between different users, including official control, requires further discussions.

References 1.

2. 3. 4.

5.

6.

Esslinger, S.; Riedl, J.; Fauhl-Hassek, C. Potential and limitations of nontargeted fingerprinting for authentication of food in official control. Food Res. Int. 2014, 60, 189–204. MoniQA. Food Authenticity - General considerations; 2013; URL http:// www.moniqa.eu/authenticity (date accessed: June 2014). Patz, C. D.; Blieke, A.; Ristow, R.; Dietrich, H. Application of FT-MIR spectrometry in wine analysis. Anal. Chim. Acta 2004, 513, 81–89. Schlesier, K.; Fauhl-Hassek, C.; Forina, M.; Cotea, V.; Kocsi, E.; Schoula, R.; van Jaarsveld, F.; Wittkowski, R. Characterisation and determination of the geographical origin of wines. Part I: overview. Eur. Food Res. Technol. 2009, 230, 1–13. Esslinger, S.; Lahrssen-Wiederholt, M.; Fauhl-Hassek, C. Fingerprinting: An innovative approach in the wine analysis [Fingerprinting: Ein innovativer Ansatz in der Weinanalytik]. Dtsch. Lebensm.-Rundsch. 2012, 108, 457–462. Roessner, U.; Nahid, A.; Chapman, B.; Hunter, A.; Bellgrad, M. Metabolomics - The Combination of Analytical Biochemistry, Biology, and Informatics. In Comprehensive Biotechnology; Scientific Fundamentals of 104 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

7.

8.

9.


10. 11. 12.

13.

14.

15.

16.

17.

18.

19.

Biotechnology; Murray, M.-Y., Ed.; Elsevier: Amsterdam, 2011; Vol. 1, pp 447−459. Eriksson, L.; Johansson, E.; Kettaneh-Wold, N.; Trygg, J.; Wikström, C.; Wold, S. Multi- and Megavariate Data Analysis Part I Basic Principles and Applications; Umetrics AB: Umeå, SW 2006. Emwas, A.-H.; Salek, R.; Griffin, J.; Merzaban, J. NMR-based metabolomics in human disease diagnosis: applications, limitations, and recommendations. Metabolomics 2013, 9, 1048–1072. Duarte, I. F.; Diaz, S. O.; Gil, A. M. NMR metabolomics of human blood and urine in disease research. J. Pharm. Biomed. Anal. 2014, 93, 17–26. Fischman, A. J.; Tompkins, R. G. Perspective on the applications of NMR in nutrition. Trends Food Sci. Technol. 1992, 3, 220–225. Rezzi, S.; Ramadan, Z.; Fay, L. B.; Kochhar, S. Nutritional Metabonomics: Applications and Perspectives. J. Proteome Res. 2007, 6, 513–525. Spyros, A.; Dais, P. NMR Spectroscopy in Food Analysis; RSC Food Analysis Monographs No. 10; Royal Society of Chemistry: Cambridge, U.K., 2013; pp 1−329. European Union. Commission Regulation (EC) No. 555/2008 laying down detailed rules for implementing Council Regulation (EC) No. 479/2008 on the common organisation of the market in wine as regards support programmes, trade with third countries, production potential and on controls in the wine sector. Off. J. Eur. Union L 2008, 170, 1–80. OIV. International Organisation of Vine and Wine. Compendium of International Methods of Analysis of Wines and Musts; 2013; URL http://www.oiv.int/oiv/info/enmethodesinternationalesvin (date accessed: June 2014). Rapp, A.; Markowetz, A.; Spraul, M.; Humpfer, E. Application of 13C-NMRSpectroscopy in wine analysis [Anwendung der 13C-NMR-Spektroskopie in der Weinanalytik]. Fresenius’ Z. Anal. Chem. 1988, 330, 462–463. Rapp, A.; Markowetz, A.; Spraul, M.; Humpfer, E. Detection and quantitative determination of sugar, sugar alcohols and sugar acids in wine by 13C-NMR-spectroscopy [Nachweis und quantitative Bestimmung von Zuckern, Zuckeralkoholen und Zuckersäuren im Wein mit 13C-NMRSpektroskopie]. Z. Lebens.-Unters. Forsch. 1989, 188, 138–143. Rapp, A.; Spraul, M.; Humpfer, E. Determination of diethylene glycol in wine with 13C NMR-spectroscopy [13C NMR-spektroskopische Bestimmung von Diethylenglykol im Wein]. Z. Lebens.-Unters. Forsch. 1986, 182, 419–421. Fauhl, C.; Wittkowski, R. Hochauflösende Kernresonanz-Spektroskopie (NMR). In Schnellmethoden zur Beurteilung von Lebensmitteln und ihren Rohstoffen; Baltes, W., Kroh, L. W., Eds.; Behr’s Verlag: Hamburg, 2004; pp 229−250. Ross, A.; Schlotterbeck, G.; Dieterle, F.; Senn, H. NMR Spectroscopy Techniques for Application to Metabonomics. In The Handbook of Metabonomics and Metabolomics; Lindon, J. C., Nicholson, J. K., Holmes, E., Eds.; Elsevier Science B.V.: Amsterdam, 2007; Chapter 3, pp 55−112. 105 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.


20. Monakhova, Y.; Schütz, B.; Schäfer, H.; Spraul, M.; Kuballa, T.; Hahn, H.; Lachenmeier, D. Validation studies for multicomponent quantitative NMR analysis: the example of apple fruit juice. Accredit. Qual. Assur. 2014, 19, 17–29. 21. Monakhova, Y. B.; Kuballa, T.; Lachenmeier, D. W. Chemometric methods in NMR spectroscopic analysis of food products. J. Anal. Chem. 2013, 68, 755–766. 22. Lachenmeier, D. W.; Humpfer, E.; Fang, F.; Schütz, B.; Dvortsak, P.; Sproll, C.; Spraul, M. NMR-Spectroscopy for Nontargeted Screening and Simultaneous Quantification of Health-Relevant Compounds in Foods: The Example of Melamine. J. Agric. Food Chem. 2009, 57, 7194–7199. 23. Koda, M.; Furihata, K.; Wei, F.; Miyakawa, T.; Tanokura, M. NMR-Based Metabolic Profiling of Rice Wines by F 2-Selective Total Correlation Spectra. J. Agric. Food Chem. 2012, 60, 4818–4825. 24. Larsen, F. H.; van den Berg, F.; Engelsen, S. B. An exploratory chemometric study of 1H NMR spectra of table wines. J. Chemom. 2006, 20, 198–208. 25. Papotti, G.; Bertelli, D.; Graziosi, R.; Silvestri, M.; Bertacchini, L.; Durante, C.; Plessi, M. Application of One- and Two-Dimensional NMR Spectroscopy for the Characterization of Protected Designation of Origin Lambrusco Wines of Modena. J. Agric. Food Chem. 2013, 61, 1741–1746. 26. Son, H. S.; Hwang, G. S.; Ahn, H. J.; Park, W. M.; Lee, C. H.; Hong, Y. S. Characterization of wines from grape varieties through multivariate statistical analysis of 1H NMR spectroscopic data. Food Res. Int. 2009, 42, 1483–1491. 27. Choi, Y. H.; Choi, H.-K.; Peltenburg-Looman, A. M. G.; Lefeber, A. W. M.; Verpoorte, R. Quantitative analysis of ginkgolic acids from Ginkgo leaves and products using 1H-NMR. Phytochem. Anal. 2004, 15, 325–330. 28. Kim, H. K.; Choi, Y. H.; Chang, W.-T.; Verpoorte, R. Quantitative analysis of ephedrine analogues from Ephedra species using 1H-NMR. Chem. Pharm. Bull. 2003, 51, 1382–1385. 29. Jiao, P.; Jia, Q.; Randel, G.; Diehl, B.; Weaver, S.; Milligan, G. Quantitative 1H-NMR spectrometry method for quality control of Aloe vera products. J. AOAC Int. 2010, 93, 842–848. 30. BioSpin, B. JuiceScreener - Enabling SGF Profiling™; 2012; URL http:// www.bruker-biospin.com/juicescreener-dir.html (date accessed: June 2014). 31. Godelmann, R.; Fang, F.; Humpfer, E.; Schütz, B.; Bansbach, M.; Schäfer, H.; Spraul, M. Targeted and Nontargeted Wine Analysis by 1H NMR Spectroscopy Combined with Multivariate Statistical Analysis. Differentiation of Important Parameters: Grape Variety, Geographical Origin, Year of Vintage. J. Agric. Food Chem. 2013, 61, 5610–5619. 32. Spraul, M.; Schütz, B.; Rinke, P.; Koswig, S.; Humpfer, E.; Schäfer, H.; Mörtter, M.; Fang, F.; Marx, U.; Minoja, A. NMR-based multi parametric quality control of fruit juices: SGF profiling. Nutrients 2009, 1, 148–155. 33. SGF. SGF International e.V. SGF Profiling; (2012; URL http://www.sgf.org/ en/home/forschung-und-projekte/profiling/ (date accessed: June 2014). 34. Monakhova, Y. B.; Schäfer, H.; Humpfer, E.; Spraul, M.; Kuballa, T.; Lachenmeier, D. W. Application of automated eightfold suppression of 106 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

35.

36. 37.


38. 39.

40. 41. 42.

43.

44.

45.

46.

47.

48.

49.

water and ethanol signals in 1H NMR to provide sensitivity for analyzing alcoholic beverages. Magn. Reson. Chem. 2011, 49, 734–739. Monakhova, Y. B.; Mushtakova, S. P.; Kuballa, T.; Lachenmeier, D. W. Investigation into the structural composition of hydroalcoholic solutions as basis for the development of multiple suppression pulse sequences for NMR measurement of alcoholic beverages. Magn. Reson. Chem. 2014 DOI:10.1002/mrc.4129. Savorani, F.; Tomasi, G.; Engelsen, S. B. icoshift: A versatile tool for the rapid alignment of 1D NMR spectra. J. Magn. Reson. 2010, 202, 190–202. Minoja, A. P.; Napoli, C. NMR screening in the quality control of food and nutraceuticals. Food Res. Int. 2014, 63 (Part B), 126–131. Wider, G.; Dreier, L. Measuring Protein Concentrations by NMR Spectroscopy. J. Am. Chem. Soc. 2006, 128, 2571–2576. Bruker. Non-Targeted Analysis/Verification; 2014; URL http://www.bruker.com/products/mr/nmr/food-screener/winescreener/ applications.html (date accessed: October 2014). Miller, J. N.; Miller, J. C. Statistics and Chemometrics for Analytical Chemistry, 4th ed.; Prentice Hall: Harlow, U.K., 2000; pp 1−271. Fauhl, C. Measurement uncertainty in wine appreciation. Mitt. Klosterneuburg 2006, 56, 3–13. Esbensen, K. H.; Guyot, D.; Westad, F.; Houmøller, L. P. Multivariate Data Analysis - in Practice : an Introduction to Multivariate Data Analysis and Experimental Design; CAMO: Oslo, 2002; pp 1−598. Massart, D. L.; Vandeginste, B. G. M.; Deming, S. N.; Michotte, Y.; Kaufman, L. Chemometrics: A Textbook; Elsevier: Amsterdam, 1988; pp 1−488. Revilla, E.; Garcı́a-Beneytez, E.; Cabello, F.; Martı́n-Ortega, G.; Ryan, J.-M. Value of high-performance liquid chromatographic analysis of anthocyanins in the differentiation of red grape cultivars and red wines made from them. J. Chromatogr. A 2001, 915, 53–60. Holbach, B.; Marx, R.; Zimmer, M. Bedeutung der Shikimisäure und des Anthocyanspektrums für die Charakterisierung von Rebsorten. Lebensmittelchemie 2001, 55, 32–34. Asenstorfer, R. E.; Jones, G. P.; Laurence, G.; Zoppi, U. Authentication of Red Wine Vintage Using Bomb-Pulse 14C. In Progress in Authentication of Food and Wine; Ebeler, S. E., Takeoka, G. R., Winterhalter, P., Eds.; ACS Symposium Series 1081; American Chemical Society: Washington, DC, 2011; Chapter 6, pp 89−99. Spraul, M. Uncovering the truth: wine analysis with next generation NMR screening. The Resonance; 2014; URL http://www.theresonance.com/ (date accessed: June 2014). Heintz, L.; Godelmann, R.; Cannet, C.; Fang, F.; Humpfer, E.; Schuetz, B.; Schäfer, H.; Spraul, M. Differentiation of Wine Grape Varieties Using 1HNMR Profiling. The Resonance; 2013; URL http://www.theresonance.com/ (date accessed: June 2014). Nietner, T.; Pfister, M.; Glomb, M. A.; Fauhl-Hassek, C. Authentication of the Botanical and Geographical Origin of Distillers Dried Grains and 107 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

50.

51. 52.


53.

54. 55.

56.

57. 58.

Solubles (DDGS) by FT-IR Spectroscopy. J. Agric. Food Chem. 2013, 61, 7225–7233. ISO International Organization for Standardization. ISO/IEC 17025:2005 General requirements for the competence of testing and calibration laboratories; 2005. MSI. The Metabolomics Standards Initiative (MSI); 2014; URL http://msiworkgroups.sourceforge.net/ (date accessed: October 2014). Naz, S.; Vallejo, M.; García, A.; Barbas, C. Method validation strategies involved in non-targeted metabolomics. J. Chromatogr. A 2014, 1353, 99–105. Zhang, L.; Nie, L. Discrimination of geographical origin and adulteration of radix astragali using fourier transform infrared spectroscopy and chemometric methods. Phytochem. Anal. 2010, 21, 609–615. Brereton, R. G. Chemometrics: Data Analysis for the Laboratory and Chemical Plant, John Wiley & Sons Ltd: Chichester, U.K., 2003; pp 1−489. Charlton, A. J.; Robb, P.; Donarski, J. A.; Godward, J. Non-targeted detection of chemical contamination in carbonated soft drinks using NMR spectroscopy, variable selection and chemometrics. Anal. Chim. Acta 2008, 618, 196–203. Haughey, S. A.; Graham, S. F.; Cancouët, E.; Elliott, C. T. The application of Near-Infrared Reflectance Spectroscopy (NIRS) to detect melamine adulteration of soya bean meal. Food Chem. 2013, 136, 1557–1561. Olivieri, A. C. Analytical Advantages of Multivariate Data Processing. One, Two, Three, Infinity? Anal. Chem. 2008, 80, 5713–5720. Trygg, J.; Gullberg, J.; Johansson, A. I.; Jonsson, P.; Moritz, T. Chemometrics in Metabolomics - An Introduction. In Plant Metabolomics; Saito, K., Dixon, R., Willmitzer, L., Eds.; Springer: Berlin, 2006; pp 117−128.

108 In Advances in Wine Research; Ebeler, et al.; ACS Symposium Series; American Chemical Society: Washington, DC, 2015.

Authentication of Wine by 1H-NMR Spectroscopy: Opportunities and

Recommend Documents